Humans readily make inferences of the social power dynamics at play across a wide range of environments. This ability requires people to possess an underlying intuitive theory of power. We tested 3 candidate formal models as hypotheses of how people judge which of two players has more power across 30 different economic games: Relative Expected Utility (the difference in expected resources), Relative Control over Resources (difference in control over the other player's resources) and Relative Choice (the difference in the amount of options each player can choose from). Our results across 3 human experiments reveal that human power judgments are best captured by combining Relative Expected Utility and Relative Choice models as joint predictors. This finding suggests that people perceive social power by considering not only who is expected to achieve their desired outcomes but also the extent of control each person holds within their environment.